The era of artificial intelligence is leading the world and has become a foremost decision-maker of human life by proving its rapid growth in every field. One of the recognized areas is super-resolution of videos or images. The world comes closer by communication means, and the visual information source has proven promising due to its ease of content understanding. So, the quality expectations during processing and after transmission for accurate decision-making grab the attention of researchers. Researchers presented a variety of techniques to improve the quality of various types of input visual data. The approaches with neural network tools are frontrunners. These solutions give rise to distinct problems concerning precision, trade-offs between accuracy and complexity, energy usage, and memory needs. The authors have addressed the benefits of modification of a basic classification neural network into regression one through experimentation particularly for super-resolution application as the neural networks basically designed for classification and not for regression purpose. The satisfactory results of performance measurement parameters, training accuracy, and root mean square error of experimentation motivate the authors for future embedding of the explored concept with a variety of data for more generalization of algorithm.